Most health apps build their engagement strategy around a reminder and a reward. Send a push at a sensible time, hand out points or a streak when the user complies, and hope the behavior sticks. It rarely does. The reward fades the moment you stop paying, and the reminder habituates into background noise.
A four-arm randomized controlled trial out of Arizona State University tested a more specific question. If you're going to spend behavioral design effort somewhere, where does it actually pay off: the reminder, the incentive, or the structure of when the incentive fires?
The answer they found is worth sitting with. The reminder did the least. The structure did the most.
The trial
Stecher and colleagues, publishing in the Journal of Behavioral Medicine in 2024, ran a four-arm RCT with 161 college students who had already said they wanted to be more active. The study (registered as NCT04691596) ran for eight weeks: four weeks of active intervention, then four weeks of follow-up with the intervention withdrawn, so they could see what survived.
They tested three ingredients in different combinations:
Personalized reminders. A push notification at a chosen time, the standard tool every app already ships.
Financial incentives. Small payments for doing the activity. This came in two flavors, and the distinction turns out to be the whole point.
Anchoring. Tying the new behavior to an existing, already-stable routine. Instead of "walk at 3 PM," the instruction is closer to "walk right after your morning coffee." The coffee is the anchor. It already happens reliably, so it can carry the new behavior on its back.
The incentive distinction is the part product teams should pay attention to. A standard incentive pays you for doing the activity at all. A cue-contingent incentive pays you only if you do the activity within roughly an hour of your chosen cue. Same money, different rule. One rewards the behavior. The other rewards the behavior happening at a consistent time, anchored to something you already do.
What won
Cue-contingent financial incentives combined with anchoring produced the highest daily step counts and the greatest odds of what the authors call temporally consistent walking, meaning walking that reliably happened at the same point in the day. That held during the four-week intervention and across the full eight weeks.
The comparison that matters: cue-contingent incentives outperformed plain incentives. Paying someone for activity worked less well than paying them for activity that happened at a consistent, anchored time. The money was the same. The rule about when the money paid out changed the outcome.
This lines up with what the habit literature has been saying for a while. A habit is, mechanically, a learned association between a context cue and a behavior. Lally and colleagues' often-cited work on how habits form found that automaticity builds when a behavior is repeated in a consistent context, and that the context consistency is doing real work, not just the repetition count. Anchoring is a way to manufacture that context consistency on purpose. Pinning the walk to the coffee gives the brain a reliable cue to attach to. Paying out only when the timing matches reinforces the cue-behavior link specifically, rather than reinforcing the behavior in the abstract.
The uncomfortable finding
Here's the result that should keep product teams honest. There were no differences in intrinsic motivation or habit strength between any of the study groups, at any time point.
Read that again. The intervention that won on behavior, more steps, more temporal consistency, did not produce measurably stronger habits or more internal motivation than the others by the study's own self-report measures. The thing that moved was behavior frequency and timing. The thing that didn't move, on these measures, was the psychological machinery we usually assume the behavior is building toward.
I don't think this means anchoring "doesn't build habits." Eight weeks is short, the habit-strength instrument is a blunt self-report tool, and behavior that's temporally consistent is exactly the raw material habits are made from. But it's a useful corrective. A product team that ships an anchoring feature and then claims it's "building lasting habits" is getting ahead of this evidence. What the trial actually demonstrated is that anchoring plus cue-contingent payment changes when and how often people act. Whether that hardens into a durable, self-sustaining habit is a separate claim that this study didn't confirm.
That gap matters because the incentive came out. When the four weeks of payment ended, the cue-contingent-plus-anchoring group still showed the best temporal consistency across the full period. So something carried past the money. But the absence of a habit-strength difference is a caution against assuming the carryover is permanent.
Why the reminder underperformed
Reminders were in the mix and they were the weakest lever. This shouldn't surprise anyone who has watched their own notification tray. A reminder is a prompt that competes with every other prompt on the phone, and it habituates fast. I wrote about exactly this in an earlier piece on notification habituation: a micro-randomized trial found standard push messages lose effect when one arrived the day before, while the long-term engagement difference between notification and no-notification arms washed out.
The mechanism difference is the lesson. A reminder pushes from outside and asks the user to interrupt whatever they're doing. An anchor pulls from inside the user's existing routine. The coffee was always going to happen. The walk just rides along. One fights the user's day. The other borrows its structure.
For anyone designing in the COM-B framework, this is a clean illustration. Reminders mostly target Motivation (a nudge to want to act). Anchoring targets Opportunity by inserting the behavior into a moment that already exists, and it lowers the Capability burden because the user doesn't have to remember or decide when. You're not adding a new slot to someone's day. You're attaching to a slot that's already reliable.
What I'd build differently
A few concrete directions come out of this.
Make users pick an anchor, not a time. Most apps ask "when do you want your reminder?" The better question is "what do you already do every day, without fail, that we can attach this to?" Coffee, the commute, brushing teeth, the post-lunch lull. The onboarding flow should surface the user's existing routines and let them clip the new behavior onto one. This is a small change in the setup wizard with a different theory underneath it.
If you pay or reward, reward the timing, not just the act. The cue-contingent result is the most actionable finding here. A streak that fires for "did the thing today" is weaker than one that fires for "did the thing within the window you anchored to." If you run points, badges, or real incentives, gate them on temporal consistency rather than mere completion. You're reinforcing the cue-behavior link, which is the thing that might actually persist.
Stop overclaiming habits. The trial moved behavior without moving self-reported habit strength in eight weeks. If your product copy promises to "build lasting habits," the evidence says you should be measuring whether that's true rather than assuming it. Track temporal consistency over months, track what happens when you reduce reinforcement, and be honest internally about the difference between "the user is acting consistently while we incentivize them" and "the user has formed a habit that survives without us."
Treat reminders as scaffolding, not strategy. Reminders aren't useless. They're a low-cost prompt that can support an anchor in the early days. But this trial puts them last for a reason. If your entire behavior-change engine is a daily push, you're leaning on the weakest of the three levers tested.
The honest summary
A four-arm trial of 161 students is small, the population is young and self-selected, and the habit measures didn't move. So I'm not going to oversell it. But the core comparison is clean and the direction is consistent with the broader habit literature: anchoring a new behavior to an existing routine, and rewarding the behavior specifically when it happens at the anchored time, beat both plain incentives and reminders on actual activity.
The cheapest thing to build is a reminder. The trial suggests it's also the least effective. The more interesting design work, asking users what they already do and clipping the new behavior onto it, is harder to ship and apparently worth more.
Paper reference: Stecher, C., Chen, C.-H., Codella, J., Cloonan, S., & Hendler, J. (2024). Combining anchoring with financial incentives to increase physical activity: a randomized controlled trial among college students. Journal of Behavioral Medicine, 47(5), 751–769. https://doi.org/10.1007/s10865-024-00492-4 | Trial registration: NCT04691596
Related: Lally, P., van Jaarsveld, C. H. M., Potts, H. W. W., & Wardle, J. (2010). How are habits formed: Modelling habit formation in the real world. European Journal of Social Psychology, 40(6), 998–1009. https://doi.org/10.1002/ejsp.674
Framework: Michie, S., van Stralen, M. M., & West, R. (2011). The behaviour change wheel: A new method for characterising and designing behaviour change interventions. Implementation Science, 6, 42. https://doi.org/10.1186/1748-5908-6-42